Scaling: from 0 to 25 million users
summary
A concrete scaling path: 28M monthly users, >400B requests/month and a 1.6M req/s peak drive specific moves — introduce Varnish request coalescing, shard caches, push proactive workers to prefill hot pages, move cache off AWS to rented bare‑metal for a 10x traffic‑bill drop, and migrate data from Mongo to Postgres 1M qps benchmark for stability. Observability uses Grafana/Prometheus and lightweight APM; analytics run in ClickHouse 1.4T rows and NATS for 500k msgs/s real‑time.
AI-written summary · about
chapters (14)
- 0:00Introduction to scaling journey
- 2:29Early caching and cache stampede
- 7:48Moving to the cloud with AWS
- 10:20Switching from NoSQL to PostgreSQL
- 13:52Database migration challenges
- 17:53Caching and cache invalidation
- 20:25Migrating off the cloud to on-prem
- 23:20Cost efficiency and trade-offs
- 26:14Custom observability and monitoring
- 31:55Global caching and latency reduction
- 34:23Kubernetes and multi-region setup
- 36:30Data analytics with ClickHouse
- 37:55Real-time updates with NATS
- 39:01Key takeaways and closing
related talks
44mIntroduction to application scaling
54mGetting to 300 million requests a month with a one man AI/Dev-Ops team
26mHow We Prepped Our Infrastructure for a Super-Bowl Sized Traffic
41mScaling Observability: From Chaos to Clarity with Open Source Tools
25mScaling Data Platforms: Managing Growth from Zero to Billions of Transactions
24mSleeping soundly even when scaling your app to 200k request/seconds
43mTools and Team for 20 Billion PageView/Month
23mIt Takes a Nation of Millions to Hold Us Back